Active Magnetic Bearings (AMB) are mechatronic systems that support a rotating shaft using magnetic levitation. The standard AMB architecture includes two radial actuators and one for the axial direction. An alternative geometry with cone-shaped magnetic cores allows for a more compact layout without a dedicated axial actuator. However, this configuration reduces the axial force generation capability and requires a more complex control architecture due to the inherent coupling of the axial and radial control actions. When using decentralized control, effective handling of the coil current limitations together with the axial disturbance rejection is difficult to achieve. In this context, the present paper demonstrates the benefits of applying Offset-Free Model Predictive Control (OF-MPC) for a cone-shaped AMB system. A procedure for the overall design is presented and supported by the experimental work conducted in a scaled machine that reproduces an on-board turbo-compressor unit for an aircraft. The modeling of the system is described together with the design of the OF-MPC in all its parts: general control architecture, disturbance model and observer design, target calculation and control problem formulation. An OF-MPC variant with reduced control horizon is proposed and implemented in real time. Experimental results demonstrate that the prototype is compliant with application-specific stability requirements from the ISO 14839-3:2006 standard. In addition, experiments show that OF-MPC outperforms decentralized PID controllers in terms of axial disturbance rejection. OF-MPC yields a favorable constrained optimal control technique for cone-shaped AMBs because intrinsic coupling and current saturation are optimally handled by the controller.